GLS or WLS

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

vahidestan
Posts: 6
Joined: Sat Sep 08, 2012 10:19 am

GLS or WLS

Postby vahidestan » Sat Sep 08, 2012 10:55 am

hi
i used OLS for a time series model then i found out the problem is heteroscedasticity so i try to estimate the model with GLS or WLS but i do not know how to do it in EViews.
also which one of the independent variables should be multiplied by the either sides of the equation to make an equation with homoscedasticitic variance?

startz
Non-normality and collinearity are NOT problems!
Posts: 3796
Joined: Wed Sep 17, 2008 2:25 pm

Re: GLS or WLS

Postby startz » Sat Sep 08, 2012 11:07 am

In the equation Estimate tab choose options and use the Weights field.

vahidestan
Posts: 6
Joined: Sat Sep 08, 2012 10:19 am

Re: GLS or WLS

Postby vahidestan » Sun Sep 09, 2012 7:08 am

i had done what you have told but i do not know how to fill the fields of "TYPE" and "WEIGHT SERIES" in the window of "equation estimation>option".
"TYPE" includes variance, S.D deviation and their inverses. which one should be used?
what should be written in the box of "WEIGHT SERIES"?

startz
Non-normality and collinearity are NOT problems!
Posts: 3796
Joined: Wed Sep 17, 2008 2:25 pm

Re: GLS or WLS

Postby startz » Sun Sep 09, 2012 7:33 am

There's no automatic answer to this question. What you do depends on your model for the heteroskedasticity.

vahidestan
Posts: 6
Joined: Sat Sep 08, 2012 10:19 am

Re: GLS or WLS

Postby vahidestan » Sun Sep 09, 2012 8:00 am

firstly thanks for ur responses
what i asked depends on the relationship between variance and one of the independent variables,
how can i find the relationship
for example if var= Z . X
var= variance
Z= what i need
X= one of the coefficients


then how can i gain variance to calculate Z ?

startz
Non-normality and collinearity are NOT problems!
Posts: 3796
Joined: Wed Sep 17, 2008 2:25 pm

Re: GLS or WLS

Postby startz » Sun Sep 09, 2012 8:44 am

I'm not sure I understand what you need. But if you are saying that the variance of the errors is proportional to an independent variable Z, then choose Type variance and weight series Z.


Return to “Estimation”

Who is online

Users browsing this forum: No registered users and 2 guests